The Measles Virus Hemagglutinin β-Propeller Head β4-β5 Hydrophobic Groove Governs Functional Interactions with Nectin-4 and CD46 but Not Those with the Signaling Lymphocytic Activation Molecule

Abstract

Wild-type measles virus (MV) strains use the signaling lymphocytic activation molecule (SLAM; CD150) and the adherens junction protein nectin-4 (poliovirus receptor-like 4 [PVRL4]) as receptors. Vaccine MV strains have adapted to use ubiquitous membrane cofactor protein (MCP; CD46) in addition. Recently solved cocrystal structures of the MV attachment protein (hemagglutinin [H]) with each receptor indicate that all three bind close to a hydrophobic groove located between blades 4 and 5 (β4-β5 groove) of the H protein β-propeller head. We used this structural information to focus our analysis of the functional footprints of the three receptors on vaccine MV H. We mutagenized this protein and tested the ability of individual mutants to support cell fusion through each receptor. The results highlighted a strong overlap between the functional footprints of nectin-4 and CD46 but not those of SLAM. A soluble form of nectin-4 abolished vaccine MV entry in nectin-4- and CD46-expressing cells but only reduced entry through SLAM. Analyses of the binding kinetics of an H mutant with the three receptors revealed that a single substitution in the β4-β5 groove drastically reduced nectin-4 and CD46 binding while minimally altering SLAM binding. We also generated recombinant viruses and analyzed their infections in cells expressing individual receptors. Introduction of a single substitution into the hydrophobic pocket affected entry through both nectin-4 and CD46 but not through SLAM. Thus, while nectin-4 and CD46 interact functionally with the H protein β4-β5 hydrophobic groove, SLAM merely covers it. This has implications for vaccine and antiviral strategies.

INTRODUCTION

Measles virus (MV), which belongs to the family Paramyxoviridae within the order Mononegavirales (1), may be the most contagious aerosol-transmitted virus circulating in human populations (2). In 2010, about 139,000 people died from secondary infections due to MV-induced immune suppression (3, 4). While a worldwide vaccination campaign aiming at eradication is ongoing (5), diminishing vaccine coverage in the United States and Europe has caused a rebound of measles cases in 2011, threatening the feasibility of eradication by 2020 (6).

MV infection starts in alveolar macrophages and dendritic cells of the airways, which transport the virus to the lymphoid organs (7, 8). MV then replicates rapidly and extensively in local lymph nodes and primary lymphatic organs. Infected immune cells circulate through the body and can transmit the infection to the respiratory epithelium, from where the virus is shed through coughing and sneezing (9–12).

Wild-type MV sequentially infects immune cells through the signaling lymphocytic activation molecule (SLAM; CD150) (13) and epithelial cells through the adherens junction protein nectin-4, also known as poliovirus receptor-like 4 (PVRL4) (12, 14, 15). In addition, the MV vaccine strain uses the ubiquitous regulator of complement activation membrane cofactor protein (MCP; CD46) as a receptor (16, 17). MV attachment to its three receptors occurs through hemagglutinin (H), a type II transmembrane glycoprotein (18). The H ectodomain is composed of a tetrameric membrane-proximal stalk topped by four six-bladed β-propeller globular heads (19). Receptor attachment to the H globular head can trigger refolding of the trimeric fusion (F) protein and cell entry.

While H-to-F signal transmission is only beginning to be understood (20–24), the molecular basis of the interactions of the receptors with H is better characterized. Initial functional analyses identified different H residues important for the interactions with the three receptors (11, 25–31). These studies were recently complemented by crystal structures of H in complex with CD46, SLAM, and nectin-4 (32–34). The SLAM (32) and CD46 (33) costructures are based on slightly different vaccine lineage H proteins, while the nectin-4 costructure (34) is based on a wild-type H protein differing by more than 17 residues from the other two proteins. Comparative analyses of the three costructures (34) concluded that a hydrophobic pocket centered in the β4-β5 groove is involved in binding of all receptors. On the other hand, previous functional analyses revealed differences between the interactions of H and individual receptors (11, 25–31).

Based on the recent identification of nectin-4 as an epithelial receptor, based on the availability of new cell lines and specific reagents for all three receptors, and focusing on the overlap of the structural footprints of the receptors, we reexamined here the functional interactions of H with its three cellular partners. We used a vaccine lineage H protein, allowing direct comparison of the interactions with CD46, SLAM, and nectin-4. We were particularly interested in functionally characterizing the interactions with nectin-4 and CD46, since separating these is necessary to generate a MV vaccine strain that cannot exit the host. Our analyses revealed a wide functional overlap between the CD46 and nectin-4 interaction sites, while the SLAM binding site is separated.

MATERIALS AND METHODS

Cells.

Vero and VerohSLAM (kindly provided by Y. Yanagi [35]) cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (HyClone, South Logan, UT) supplemented with 10% fetal bovine serum (FBS). G418 at a concentration of 0.5 mg/ml was added to the medium of VerohSLAM cells. Cells were kept at 37°C with 5% CO₂. Chinese hamster ovary (CHO) cells were maintained in RPMI medium (Corning, Manassas, VA) supplemented with 10% FBS and 0.5 mg/ml of nonessential amino acids. CHO-N4 cells (expressing human nectin-4, kindly provided by M. Lopez [36]) and CHO-SLAM cells (expressing human SLAM) were grown in RPMI medium complemented with 0.5 mg/ml of G418; CHO-CD46 cells (expressing human CD46) were grown in RPMI medium complemented with 1 mg/ml of zeocin. Baby hamster kidney cells expressing stably the T7 polymerase (BHK-T7) (37) were maintained in DMEM supplemented with 10% FBS and 0.1 mg/ml G418.

Plasmids and mutagenesis.

The H and F open reading frames (ORFs) of wild-type MV (wt IC-B 323 [38]) or vaccine MV (39) were cloned into a pCG vector (40). Mutations were introduced into the pCG-Hvac plasmid by using the QuikChange site-directed mutagenesis (Stratagene) protocol according to the manufacturer's instructions. Introduction of the desired mutations was confirmed by sequencing.

Fusion assays.

For the semiquantitative fusion experiments, cells were plated at 80% confluence prior to transfection. Equal amounts of pCG-F and the mutated pCG-H were transfected by using Lipofectamine 2000 (Life Technologies, Grand Island, NY) according to the manufacturer's instructions. Fusion scores were determined at 24 to 48 h posttransfection according to the extent of syncytium formation in a field of view representative of the well. A syncytium was defined as a cell with 3 or more nuclei. In CHO-N4 cells, the extent of syncytium formation was limited compared to that in other cell lines, and the number of nuclei per syncytium rarely exceeded 10 nuclei. Thus, in this cell line, the number of syncytia per field of view was counted. A fusion score of 3 was attributed to the positive control H^vac2. Setting the positive control at 100% fusion, a fusion score of 3 was attributed to H proteins showing more than 75% fusion, a fusion score of 2 denoted fusion averaging 50 to 75%, a fusion score of 1 was attributed to mutants showing between 25 and 50% fusion, and a fusion score of 0 denoted <25% fusion. In CHO-CD46, CHO-SLAM, Vero, and VerohSLAM cells, a fusion score of 3 denotes wild-type fusion levels (on average more than 15 nuclei per syncytium), 2 denotes 5 to 15 nuclei per syncytium, 1 denotes 3 to 5 nuclei per syncytium, and 0 denotes <3 nuclei per syncytium. An average fusion score was assigned for each mutant after at least three independent experiments.

For quantitative fusion assays, BHK-T7 cells in 24-well plates were transfected with equal amounts (0.5 μg) of pCG-F and the mutated pCG-H using Lipofectamine 2000 according to the manufacturer's protocol. CHO-N4, CHO-CD46, or CHO-SLAM cells in 6-well plates were transfected with 1.4 μg of pTM1-luc. Six hours after transfection, the CHO cells were washed and detached by using Versene (Gibco by Life Technologies, Grand Island, NY) and then overlaid onto the BHK-T7 cells. After 16 h, the expression of luciferase was quantified by using the Steady-Glo luciferase activity system (Promega Corporation, Madison, WI) and a Topcount NXT luminometer (Packard Instrument Company, Meriden, CT).

Surface plasmon resonance.

The soluble H protein ectodomain carrying the Y543A substitution was expressed and purified as described previously (27). The receptor binding kinetics of the mutant H protein was determined essentially as described previously (27). Briefly, the interaction of the soluble H protein ectodomain with the relevant soluble domains of the three receptors SLAM, CD46, and nectin-4 was monitored by using a BIAcore 3000 instrument (GE Healthcare, Pittsburg, PA). HBS-EP (GE Healthcare, Pittsburg, PA) buffer (0.01 M HEPES [pH 7.4], 0.15 M NaCl, 3 mM EDTA, 0.005% [vol/vol] surfactant P20) was used for all protein dilutions and as the running buffer. The FLAG-tagged H protein was captured onto the CM5 sensor chip surface by an immobilized anti-FLAG antibody (Sigma, St. Louis, MO). To monitor receptor binding, different concentrations of soluble receptors (0.05 μM to 0.8 μM) were injected over the captured MV H proteins as indicated. The soluble receptors were injected over the CM5 sensor chip surfaces either exhibiting or lacking captured H protein. The surfaces lacking captured H served as control surfaces for nonspecific receptor binding. The antibody surface was regenerated by injection of 50 mM citric acid (pH 3.0) for 60 s followed by 10 mM glycine (pH 3.0) for 60 s, with little change in the activity of the monoclonal antibody surface monitored. The curve-fitting function of BIAevaluation 4.1 software was used to fit rate equations derived from the 1:1 Langmuir binding model to the experimental data. The equilibrium dissociation constant (K_D) was determined from the kinetic rate constants K_on (association) and K_off (dissociation).

Viruses and infections.

To generate MV^vac2L482S and MV^vac2Y543A, the H ORFs were transferred into an infectious MV genome encoding green fluorescent protein (GFP) upstream of N [p(+)MV^vac2eGFP], using the PacI and SpeI restriction sites (39). MV^vac2Luc expressing luciferase was generated as described previously (41). Briefly, the luciferase ORF was cloned into the infectious MV genome by using the MluI and AatII restriction sites upstream of N. Recombinant viruses were rescued as described previously (42). To prepare virus stocks, VerohSLAM cells were infected at a multiplicity of infection (MOI) of 0.03 and then incubated at 32°C for 2 to 3 days. To harvest virus, cells were scraped into Opti-MEM I reduced-serum medium and freeze-thawed twice. Titers were determined by 50% tissue culture infective dose (TCID₅₀) titration on VerohSLAM cells. For infections with the GFP-expressing viruses, cells were infected at the indicated MOIs for 1 h at 37°C in Opti-MEM I reduced-serum medium. Cells were then washed and incubated in 5% FBS medium at 37°C with 5% CO₂ for the indicated times. Fluorescence microscopy photographs were taken at the indicated times postinfection. For infections with MV^vac2Luc, 5 × 10⁴ cells were infected with 10,000 TCID₅₀ in the presence of 5 μg/ml of a CD46 blocking antibody (M177, a kind gift of D. Gerlier and T. Seya), a nectin-4 blocking antibody (N4.61, kindly provided by M. Lopez), or a SLAM blocking antibody (IPO-3; Abcam, Cambridge, MA) or with 10 μg/ml of soluble protein N4VCC-Fc and the negative control cytotoxic-T lymphocyte-associated antigen 4 (CTLA-4)–Fc (kindly provided by M. Lopez). Luciferase expression was determined at 16 h postinfection by using the Steady-Glo luciferase activity system (Promega Corporation, Madison, WI) and a Topcount NXT luminometer (Packard Instrument Company, Meriden, CT).

RESULTS

Functional analysis of MV vaccine H interactions with different receptors.

The recent availability of cocrystal structures of MV-H in complex with CD46 (33) and SLAM (32) allowed a focused reexamination of the functional interactions of this protein. Figure 1 illustrates the MV-H residues contacting the different receptors. The CD46 contact residues (Fig. 1A, red residues) are situated on propeller blades β3, β4, and β5 (Fig. 1D, red, orange, and yellow backbone segments, respectively). The SLAM contact residues localize primarily in propeller blades β5 and β6 (Fig. 1C, blue residues).

Fig 1.

Receptor footprints on the MV H protein. Side views of the six-bladed β-propeller sheet H protein head crystal structure (Protein Data Bank [PDB] accession number 3INB) as a ribbon plot. (A) CD46 contact residues (33) (red). (B) Residues important for the nectin-4 functional interactions (11, 30) (green). (C) SLAM contact residues (32) (blue). (D) Overlap between the footprints, with 3 residues (F483, Y541, and Y543) relevant for the interactions with all three receptors shown in orange. The backbones of propeller blades β3, β4, and β5 are shown in red, orange, and yellow, respectively.

When this study was planned, the costructure of MV-H with nectin-4 was not available, but 5 residues functionally important for this interaction had been mapped (Fig. 1B, green residues). Interestingly, these analyses indicated that the residues relevant for the nectin-4 functional interaction coincide with three CD46 and SLAM contact residues: F483, Y541, and Y543 (highlighted in orange in Fig. 1D). As these observations suggested the existence of a functional overlap between the three receptor binding sites, we set out to characterize its extent.

To directly compare the functional interactions of all three receptors, we introduced amino acid changes in the vaccine lineage H^vac2 protein (39). We focused on mutations expected to discriminate between the receptors: I194S, D530A, and R533A, known to be SLAM specific; V451S, L464S, Y481A, K488A, T498A, and L500S, thought to be CD46 specific, and L482S and P497S, thought to be nectin-4 specific. These mutations are listed in numerical order in Fig. 2A. We also reanalyzed the receptor-dependent fusion function of F483S, Y541A, and Y543A, known to be in close proximity to all three receptors.

Fig 2.

Receptor-dependent fusion efficiency of H^vac2 mutants. (A) Fusion activity of H^vac2 mutants in cells expressing CD46, nectin-4, or SLAM. Fusion levels, measured as described in Materials and Methods, were scored 0 to 3. (B) Quantitative fusion assay of selected H^vac2 mutants. Means and standard deviations of the fusion activity of each mutant are shown as percentiles of the positive control H^vac2. Significant differences among the mutants (P < 0.05 by t test) are indicated with an asterisk; n.s. indicates nonsignificant differences (P > 0.05).

The first round of analysis (Fig. 2A) was based on a semiquantitative fusion assay and included the 14 mutants listed above. Control fusion assays with wild-type or vaccine H proteins served as standards for quantification (Fig. 2A, top). The I194S, D530A, and R533A substitutions abolished SLAM usage without drastically affecting fusion through CD46 and nectin-4, as expected. On the other hand, of the 11 other mutations, 9 strongly affected both CD46 and nectin-4 interactions. The exceptions were L482S, which affected fusion only through nectin-4, and L500S, which affected fusion only through CD46. Therefore, these analyses revealed strikingly similar functional interactions of H^vac2 with CD46 and nectin-4.

The second round of analysis focused on the critical nectin-4 residues L482, F483, P497, Y541, and Y543 as well as on the critical CD46 residue Y481. This analysis was based on a quantitative fusion assay. As with the semiquantitative assay, SLAM-dependent fusion was at most marginally affected by the six mutations (Fig. 2B, blue columns). However, all six mutations affected fusion through CD46 and nectin-4, with F483S, Y541A, and Y543A almost completely abolishing fusion through both receptors (Fig. 2B, compare red and green columns). We noted occasional differences between the results of the semiquantitative assays and those of the quantitative assays. However, these differences were small. The Y481A substitution reduced fusion to the level of the negative control in CD46-expressing cells and also reduced fusion through nectin-4 by about 65% compared to H^vac2. Thus, the results of both fusion assays consistently revealed striking similarities of the modes of H binding to CD46 and nectin-4. On the other hand, even if structural studies documented proximity of SLAM densities to H residues F483, Y541, and Y543, our SLAM-dependent fusion assays did not detect functional relevance.

Competition between nectin-4- and CD46-dependent entry.

We next tested whether reagents blocking entry through one MV receptor interfered with entry through the others (Fig. 3). We used CHO cells expressing only CD46, nectin-4, or SLAM. We first infected cells treated with antibodies known to block entry through each receptor. As expected, CD46 antibody M177 reduced entry and replication of MV^vac2 by 85% on CD46-expressing CHO cells but had no effect on entry through nectin-4 or SLAM. Similarly, nectin-4 antibody N4.61 reduced virus entry by 95% on CHO cells expressing nectin-4. Finally, SLAM antibody IPO-3 reduced virus entry by 80% on SLAM-expressing CHO cells. The latter two antibodies had little to no effect on entry through the other receptors.

Fig 3.

Receptor-dependent cell entry in the presence of antibodies or soluble nectin-4. CHO cells expressing CD46, nectin-4, or SLAM were infected with MV^vac2Luc at an MOI of 0.2 in the absence of treatment (NT, not treated) or in the presence of CD46 blocking antibody M177 (anti-CD46), nectin-4 blocking antibody N4.61 (anti-nectin-4), SLAM blocking antibody IPO-3 (anti-SLAM), soluble nectin-4 ectodomain N4VCC (sN4), or soluble CTLA4 (sCTLA4) as a negative control. Another control consisted of noninfected cells (no virus). At 16 h postinfection, cells were analyzed for luciferase expression levels.

We then assessed whether addition of the N4VCC nectin-4 ectodomain to the medium of receptor-expressing CHO cells interfered with MV entry. In cells expressing CD46 or nectin-4, N4VCC reduced entry and replication of MV^vac2 by 93% and 95%, respectively, but infection of SLAM-expressing CHO cells was only halved. The control soluble protein CTLA-4 had marginal or no effect on nectin-4-, CD46-, and SLAM-dependent entry, as expected. Thus, soluble nectin-4 can bind to MV H and interfere with CD46 receptor recognition. However, competition of soluble nectin-4 with SLAM-dependent entry is less effective. We also performed competition experiments using monomeric soluble CD46, but even at a high concentration, competition was not observed. This is in agreement with previous studies documenting that oligomerization of CD46 is necessary to sustain MV neutralization (43).

The Y543A mutation strongly reduces H binding to nectin-4 and CD46 while leaving SLAM binding unaffected.

With a biochemical approach, we generated a mutated H protein soluble ectodomain and measured its kinetics of binding to the relevant domains of the three receptors. We focused on the Y543A mutation because this is the substitution with the clearest differential effect on nectin-4, CD46, and SLAM membrane fusion functions. Interestingly, the aromatic ring of tyrosine 543 is located within 4.5 Å of residues of all three receptors in the cocrystals (34). The H–nectin-4, -CD46 and -SLAM equilibrium dissociation constants (K_D) determined previously were 20, 79, and 80 nM, respectively (Table 1) (12, 27). The Y543A mutation reduced the affinity of H for nectin-4 and CD46 about 65 and 25 times to K_D values of 1.3 μM and 2.1 μM, respectively (Table 1). A drop in the association rate (K_on) was mainly responsible for the decreased K_D. On the other hand, the H^Y543A-SLAM affinity was only about 2-fold reduced to 189 nM (Fig. 4C). Thus, Y543 is critical for H binding to both nectin-4 and CD46 but not to SLAM.

Table 1.

Kinetics of binding of soluble H and H^Y543A ectodomains to the relevant domains of the three MV receptors

Analyte	Ligand	Kon (103) (M−1 s−1)	Koff (10−3) (s−1)	KD (Koff/Kon)
H	CD46a	27	2.1	79 nM
	Nectin-4b	176	3.5	20 nM
	SLAMa	25	2	80 nM
HY543A	CD46	2.6	5.6	2.1 μM
	Nectin-4	5.5	7.3	1.3 μM
	SLAM	31	5.8	189 nM

^aData are from reference 27.

^bData are from reference 12.

Fig 4.

Generation and characterization of two viruses with receptor-restricted cell fusion. (A) Schematic representation of the viral genomes of MV^vac2, MV^vac2L482S, and MV^vac2Y543A encoding GFP from an additional open reading frame upstream of the N gene. (B) Comparative analysis of the spread of MV^vac2 and MV^vac2Y543A in cells expressing a single receptor. Cells were infected at an MOI of 0.1, and the infection was documented by microscopic observation of the GFP signal after 3 days. (C) Comparative analysis of the spread of MV^vac2 and MV^vac2L482S in cells expressing a single receptor. Cells were infected at an MOI of 0.5, and the infection was monitored by microscopic observation of the GFP signal during the following 3 days. Note that in CHO-SLAM cells, syncytia form very rapidly, and most cells are lysed between days 2 and 3, accounting for the decrease in GFP fluorescence. (D) Titers of cell-associated virus (cell fraction) and cell-free virus (supernatant) produced following infection of CHO-N4 cells with MV^vac2 and MV^vac2L482S.

Generation of two viruses with receptor-restricted cell fusion.

We then generated a recombinant virus with mutated H tyrosine 543, MV^vac2Y543A (Fig. 5A). From the above-described results, we expected this virus to be unable to enter cells through CD46 and nectin-4 but to retain SLAM-dependent entry. We also attempted to generate a MV retaining entry through CD46 and SLAM but not through nectin-4. Such a vaccine lineage virus could be used in the final phases of MV eradication, as it would not be shed. Since the L482S substitution drastically reduced fusion through nectin-4 without affecting fusion through CD46 and SLAM (Fig. 2A), we generated MV^vac2L482S (Fig. 4A).

Fig 5.

H residues relevant for functional interactions with the three receptors. (A to E) Side view of the six-bladed β-propeller sheet H protein head crystal structure in a gray surface representation. The three receptors are shown in panels C to E as ribbon plots. Green, nectin-4; red, CD46; blue, SLAM. (F to H) The same structures rotated 60° backwards compared to the complexes in panels C to E. (I to K) Detail of how the three receptors either penetrate or cover the hydrophobic groove, in the same orientation as panels F to H. (A) Residues of the hydrophobic groove shown here to be functionally important for the interaction with nectin-4 and CD46 are shown in orange. Residue Y543 is in magenta, residue L482 is in green, residue L500 is in red, and residues I194, D530, and R533 are in blue. (B) Hydrophobic groove residues proposed to be important based on structural data are shown in yellow. (C, F, and I) Binding of the nectin-4 IgV domain (green). (D, G, and J) Binding of the two CD46 membrane-distal domains (red). (E, H, and K) Binding of the SLAM IgV domain (blue). The amino-terminal residue of each receptor is shown in black. Structural coordinates are from PDB accession numbers 3INB for panels A, B, D, G, and J; 4GJT for panels C, F, and I; and 3ALZ for panels E, H, and K. The PDB files were aligned before comparison. All panels were generated by using Pymol (http://www.pymol.org/).

Both MV^vac2Y543A and MV^vac2L482S were rescued and amplified on VerohSLAM cells expressing both SLAM and CD46, on which they reached high titers and induced syncytium formation to the same extent as MV^vac2 (data not shown). Their replication was then analyzed on CHO cells expressing a single receptor: nectin-4, CD46, or SLAM. Whereas MV^vac2 replicated and induced syncytium formation in the three cell lines, MV^vac2Y543A did so only in cells expressing SLAM (Fig. 4B). The latter observation confirmed the importance of residue Y543 for entry through both nectin-4 and CD46 but not through SLAM.

As expected, MV^vac2L482S replicated approximately as efficiently as MV^vac2 in Vero cells expressing CD46 and in CHO cells expressing SLAM, as monitored by immunofluorescence (Fig. 4C). Surprisingly, it also entered and replicated in CHO cells expressing nectin-4. However, by day 2, MV^vac2 induced syncytium formation, whereas no syncytium was observed in cells infected with MV^vac2L482S at day 2 or at later time points. Despite this striking difference in cytopathic effect, MV^vac2L482S produced intracellular and supernatant infectivity with kinetics similar to those of MV^vac2 in nectin-4-expressing CHO cells (Fig. 4D). Hence, the L482S substitution does not abrogate cell entry of MV through nectin-4 but modulates the extent of subsequent cell-to-cell fusion through this receptor. The latter observation explains the null phenotype of H^vac2L482S in the fusion assay.

DISCUSSION

Vaccine strain MV H protein interacts with three receptors: CD46, SLAM, and nectin-4. Initial mapping of the functional footprints of CD46 and SLAM on H revealed adjacent interaction surfaces with some overlap (31), and recent structural analyses (32, 33) reached a similar conclusion. In particular, CD46-H contacts are located in H head propeller blades β4 and β5, where amino acids L464, L500, Y541, and Y543 form a hydrophobic groove contacting the CD46 D′D loop; residues Y481 to F483 are also important for receptor specificity (33). SLAM contacts H through salt bridges involving residues D505, D507, D530, and R533; an intermolecular β-sheet involving residues P191 to R195; and hydrophobic interactions involving residues F483, Y524, Y541, Y543, P545, and F552 (32). We confirmed here the functional importance of the salt bridges and intermolecular β-sheet in the interaction with SLAM. On the other hand, we noted that residues F483, Y541, and Y543 are not critical for SLAM recognition when mutated individually.

When we started this study, no crystal structure was available for MV H in complex with nectin-4, but two independent functional studies characterized H residues L482, Y483, P497, Y541, and Y543 as important for function through the then-unknown epithelial receptor now identified as nectin-4 (11, 30). Notably, mutations of these residues had minor or no effects on SLAM usage, suggesting that the nectin-4 and SLAM binding sites do not overlap functionally. Here, we confirmed the functional relevance of these five residues for the nectin-4 interaction in a MV vaccine background. Moreover, we documented the relevance of eight additional H protein residues in the nectin-4 interaction. Strikingly, almost every residue affecting this interaction also affected CD46 recognition but had little to no effect on the use of SLAM for fusion. Thus, the functional interactions of vaccine MV H with nectin-4 and CD46, but not with SLAM, are markedly similar.

Very recently, the crystal structure of a wild-type MV H protein in complex with nectin-4 was solved, defining 29 amino acids located <4.5 Å away from H (34). Among these residues, L464, F483, L482, L500, Y524, Y543, and S548 define a hydrophobic groove located between β-propeller blades 4 and 5 (Fig. 5B, yellow residues) (33, 34). Our functional studies identify a similar H protein surface as relevant for cell fusion through both CD46 and nectin-4. This surface is defined by residues L464 and Y483 in the β4 blade and by residues Y541 and Y543 in the β5 blade (Fig. 5A, orange and magenta residues). In addition, we show that L500 is CD46 specific (Fig. 5A, red residue), whereas L482 is functionally relevant only for the nectin-4 interaction (Fig. 5A, green residue).

Structural analogies extend to how the hydrophobic groove accommodates either the FG loop of nectin-4 (Fig. 5C, F, and I) or the D′D loop of CD46 (Fig. 5D, G, and J). In particular, tyrosine 543 (magenta) holds the CD46 D′D loop and the nectin-4 FG loop in the groove, along with L464 and Y483. This explains why the Y543A substitution drastically reduces binding, fusion, and entry through both receptors. Altogether, structural and functional data are consistent with similar modes of binding to the same hydrophobic groove for both CD46 and nectin-4.

Interestingly, Zhang et al. suggested that this hydrophobic groove is also crucial for the SLAM interaction and that an antiviral drug could block H binding to all three receptors (34). However, the SLAM functional footprint defined here does not overlap those of CD46 or nectin-4. Structurally, we note that SLAM does not penetrate the hydrophobic groove deeply like CD46 and nectin-4. Rather, its CC′ loop merely covers it (Fig. 5E, H, and K). In fact, the residues important for SLAM-mediated cell fusion are located away from the hydrophobic pocket, in H blades β5 and β6 (Fig. 5A, blue residues). Consistent with this, a soluble nectin-4 ectodomain strongly interfered with entry through CD46 but had much smaller effects on SLAM-dependent entry. Thus, a drug targeting the nectin-4 and CD46 binding sites is unlikely to also inhibit SLAM binding.

The original MV vaccine was attenuated by serial passage of the Edmonston isolate in different cell types (44), resulting in multiple mutations in different genes (45). Among the H gene mutations, a few facilitate CD46 recognition, with N481Y being the most critical (46). Our results suggest that the CD46 binding site may have evolved from the nectin-4 binding site; minimal changes have allowed binding to a new receptor without disrupting the nectin-4 binding function. Indeed, vaccine MV infects nectin-4-expressing cell lines with an efficiency similar to that of wild-type strains (14). Remarkably, a wild-type virus expressing the vaccine H protein still replicates preferentially if not exclusively in lymphoid (SLAM-expressing) and respiratory (nectin-4-expressing) tissues in vivo (47), implying dominance of entry through the natural receptors for MV spread.

Our attempt to generate a virus that is able to enter cells through CD46 but not nectin-4 based on the L482S substitution was not successful: this virus lost its ability to cause formation of large syncytia on nectin-4-expressing cells, but it entered these cells and replicated efficiently. In retrospect, this may not be surprising, because our screening assay is based exclusively on observations of receptor-specific syncytium formation.

Should a virus that cannot exit the host still be sought? While vaccine shedding is well documented (48–50), horizontal transmission of vaccine is usually benign. Nevertheless, virus shedding should be avoided during the final phase of MV eradication. The issue of shedding vaccinees is addressed by the use of a chemically inactivated poliovirus for the respective eradication campaign (51), but efforts to develop an inactivated MV vaccine ceased after formalin treatment was associated with more severe disease (52). The availability of structural information on nectin-4 now makes possible a structure-guided approach to the identification of a set of mutations discriminating between CD46 and nectin-4 interactions and the generation of a vaccine lineage virus that cannot be shed. Screening of mutant H proteins will have to be based on both fusion assays and determination of the binding affinities for both nectin-4 and CD46.

In summary, our data confirm and refine the characterization of the β4-β5 H protein hydrophobic groove as the binding site for both CD46 and nectin-4. However, this hydrophobic groove is not relevant for entry through SLAM, which interacts functionally with propeller blades β5 and β6. Since SLAM recognition precedes nectin-4 binding during the viral life cycle, efforts to develop an antiviral drug should preferentially be directed to the SLAM interaction.